	US 11,954,417 B1
	Method for predicting permeability of multi-mineral phase digital core based on deep learning
Hai Sun, Qingdao (CN); Liang Zhou, Qingdao (CN); Dongyan Fan, Qingdao (CN); Lei Zhang, Qingdao (CN); Jun Yao, Qingdao (CN); Yongfei Yang, Qingdao (CN); Kai Zhang, Qingdao (CN); Qian Sang, Qingdao (CN); Xia Yan, Qingdao (CN); Lei Liu, Qingdao (CN); Fei Luo, Qingdao (CN); and Yuda Kan, Qingdao (CN)
Assigned to CHINA UNIVERSITY OF PETROLEUM (EAST CHINA), Qingdao (CN)
Filed by CHINA UNIVERSITY OF PETROLEUM (EAST CHINA), Qingdao (CN)
Filed on Nov. 8, 2023, as Appl. No. 18/504,557.
Claims priority of application No. 202211395425.8 (CN), filed on Nov. 9, 2022.
Int. Cl. G06F 30/28 (2020.01); G06F 30/27 (2020.01); G06F 113/08 (2020.01)

CPC G06F 30/28 (2020.01) [G06F 30/27 (2020.01); G06F 2113/08 (2020.01)]

8 Claims

1. A method for predicting permeability of a multi-mineral phase digital core based on deep learning, specifically comprising:

step 1, constructing a three-dimensional digital core and randomly generating a pore structure in the three-dimensional digital core, wherein the pore structure comprises organic matter pores and inorganic matter pores, and for the three-dimensional digital core, the pore structure is filled with a fluid and a rest thereof is set as a skeleton;

step 2, acquiring a plurality of multi-mineral digital core images by performing image segmentation on the constructed three-dimensional digital core;

step 3, acquiring permeability corresponding to each of the multi-mineral digital core images by using multi-physics field simulation software, and constructing a multi-mineral digital core data set based on the plurality of multi-mineral digital core images and the permeability corresponding to each of the multi-mineral digital core images;

step 4, constructing an SE-ResNet18 convolutional neural network, training the SE-ResNet18 convolutional neural network with the multi-mineral digital core data set, and calculating permeability corresponding to each of the multi-mineral digital core images; and

step 5, inputting an image of a multi-mineral core to be predicted into the trained SE-ResNet18 convolutional neural network, and obtaining permeability corresponding to the multi-mineral core with the trained SE-ResNet18 convolutional neural network according to the image of the multi-mineral core to be predicted, wherein

step 3 specifically comprises:

step 3.1, setting a density and viscosity of a fluid in a multi-mineral digital core, setting a side of the multi-mineral digital core as a fluid inlet, setting another side of the multi-mineral digital core as a fluid outlet, setting inlet pressure and outlet pressure of the fluid, and setting a wall slip length of the organic matter pores and a wall slip length of the inorganic matter pores in the multi-mineral digital core;

step 3.2, for each of the multi-mineral digital core images, constructing a multi-mineral digital core based on each of the multi-mineral digital core images, acquiring a flow process of the fluid in each multi-mineral digital core by simulation using the multi-physics field simulation software, acquiring a flow field distribution of each multi-mineral digital core in a stable state, and calculating permeability of each multi-mineral digital core; and

step 3.3, by taking the permeability corresponding to each of the multi-mineral digital core images as labels and the plurality of multi-mineral digital core images and the permeability corresponding to each of the multi-mineral digital core images as sample data, constructing the multi-mineral digital core data set and dividing the multi-mineral digital core data set into a training set, a test set, and a validation set.